The study was carried out during a monitoring project between May 2013 and September 2014 in the mountainous areas in Bulgaria. The field studies were conducted in 11 sampling sites in the following mountains: Rila, Pirin, Belasitsa, Vitosha, Osogovo, Western Stara Planina, Central Stara Planina, Eastern Stara Planina, Western Rhodopes, Eastern Rhodopes and Strandza (Fig. 1). In the fieldwork of the recent monitoring project, the efforts were concentrated on the most suitable habitats of RWA in places with registrations of literature data.

Figure 1. 

The eleven sampling monitoring sites (mountains): 1 Western Stara Planina, 2 Central Stara Planina, 3 Eastern Stara Planina, 4 Vitosha, 5 Osogovo, 6 Belasitsa, 7 Rila, 8 Pirin, 9 Western Rhodopes, 10 Eastern Rhodopes, 11 Strandza.

For the four ant species, sampling transects were set in areas with appropriate homogeneous biotopes. For a homogeneous biotope, a continuous polygon was taken from the potential habitat of each species, which was falling within a single monitoring area characterised by similar environmental features. Potential habitats are predetermined by mapping patterns or on the terrain. Potential habitat GIS models for each target RWA were made using the “intersect” tool of ArcGIS 10 with polygon vector layers of Bulgarian Forest GIS database: Dominant tree type, Canopy cover <80%, Forest age: young (<50 years), middle (50–100), old (>100 years) and Elevation according to the ecological preferences of each species by literature data for Bulgarian populations (Atanassov and Dlusskij 1992).

According to the Bulgarian landscape characteristics and specifics, we used a combination of sampling methods from other countries in the temperate climatic zone (Cherix 1977; Domisch et al. 2005; Cherix et al. 2007, 2012; Hughes and Broome 2007; Gotelli et al. 2011; Borkin et al. 2012; Zakharov et al. 2013, and Breen 2014).

In each selected monitoring area, a minimum of 8 sampling transects of 250 m and a width of 5 m were examined (i.e. 2 km length in total and 1250 m² per transect) across a homogeneous biotope (Borkin et al. 2012). In total, 172 sampling transects (21.5 ha) were selected. Eight to 29 transects were sampled per site (i.e. at least 10 000 m²) (Pętal and Pisarski 1966). Transects were separated by at least 10 meters to avoid counting nests twice (Leponce et al. 2004). Each transect was visited once per year during the daylight from May to the end of September. In each transect, one GPS point was taken at the beginning of it, one at its end and one point per each localised nest inside the transect (by Garmin MAP 60CS). Thus, the same transects could be used for future monitoring studies.

The separate nests (active and abandoned) were counted, with a diameter greater than 20 cm (Domisch et al. 2005; Zakharov et al. 2013). Each nest was digitally photographed, including the surrounding vegetation within a radius of up to 30 m (Cherix et al. 2007, 2012). At least 10 ant specimens were taken from each nest in 95% ethanol (Bestelmeyer et al. 2000). All the samples were preserved in the collection of V. Antonova at the IBER, Bulgarian Academy of Sciences, Sofia.

The nests’ description and environmental variables were filled in a field form (see Suppl. material 1). One field form was completed for each transect.

1. Nest number: number of nests of each species within one sampling transect (1250 m ²)
2. Nest measurements and identification:

• Diameter of the nest (± 5 cm) (for further monitoring);
• Height of the nest (± 5 cm) (vertically from the ground level to the top of the nest, for further monitoring);
• Active/abandoned: Binary variable (abandoned: for further monitoring) (Domisch et al. 2005); A mound with workers only passing on it was not considered as an active.
• Species. The determination of the species was done in a laboratory using the identification keys of Atanassov and Dlusskij (1992); Czechowski et al. (2012) and Seifert (2018).
• Cluster of colonies (more than one mound per colony): Binary variable (yes/no). It was noted whether the nest was isolated or there was a cluster formed (Hughes and Broome 2007) by observing the workers’ behaviour by „transplant experiments”: when a worker from one nest being placed near another nest is aggressively attacked by the others, the nests belong to different colonies (Kaspari 2000).

Predictor variables recorded were related to the topography and habitat characteristics of each sampling transect. Data for the first 10 variables were taken in situ as approximate assessment, based on the whole transect range:

1. Elevation: based on GPS current data.
2. Exposure of the slope (based on a compass) 8 variables: N, NW, W, SW, S, SE, E, NE.
3. Slope as degrees in 4 classes: 0–5, 6–15, 16–30, >30 degrees.
4. Ecological groups of plants (adapted for Bulgaria, Lyubenova 2004), defined by dominant plants species in respect of their adaptations to soil and air humidity in 5 categories: xerophytes (low humidity – succulents, Euphorbia falcata, Dianthus petraeus, Poa bulbosa, etc.); meso-xerophytes (drought-tolerant – Hypericum perforatum, Adonis vernalis, Potentilla argentea, Quercus pubescens, etc.); mesophytes (moderate humidity – Bellis perennis, Festuca pratensis, Medicago arabica, Alcea rosea, etc.); meso-hygrophytes (increased moisture affinity – Ranunculus sceleratus, Carex distans, Juncus bufonis, etc.); hygrophytes (high moisture – mosses, Caltha palustris, Epilobium hirsutum, Oxalis acetosella, etc.).
5. Stone cover as percentage in 5 classes: 0–5%, 6–25%, 26–50%, 51–75%, 76–100%.
6. Habitat type in 7 categories by EUNIS habitat classification (https://www.eea.europa.eu/data-and-maps/data/eunis-habitat-classification, Level 1 and 2; ecotones added): coniferous forest, ecotone of coniferous forest, broadleaved deciduous forest, ecotone of deciduous forest, mixed forest, ecotone of mixed forest, grasslands.
7. Grass cover as a percentage in 4 classes: <25%, 25–49%, 50–74%, 75–100%.
8. Canopy cover as a percentage in 5 classes: <20%, 20–40%, 41–60%, 61–80%, >80%.
9. Undergrowth (shrub) density as a percentage in 4 classes: <5%, 6–25%, 26–50%, 51–100%.
10. Dominant tree species of the forest in 12 categories: Picea abies, Pinus sylvestris, Pinus nigra, Pinus mugo, Pinus peuce/ heldreichii, Pinus spp. with Fagus spp., Juniperus spp., Fagus spp., Quercus spp., Castanea sativa, single trees (fruit trees), grass (open habitat). Identification keys according to Delipavlov et al. (2003) were used.
11. Forest mean age as years in 4 classes: 0–50, 51–100, 101–150, >150. The data were taken from the Bulgarian Forest GIS database (www.agrolesproject.com/cgi-bin/agro1?m=med6).

Nest density calculations for each monitoring site were based on the total number of nests of the particular species and the total area of its sampling transects. For each monitoring site, the study surface area (the number of transects) increased proportionally to the area of the appropriate forest habitats for RWA.

Non-parametric statistics were used as our data have not normal distribution and could not be normalised by a transformation. The species represented by < 5 mounds were not included in the statistical analyses as they were too rare. Binomial logistic regression with a logit link function was used for a detailed study of the predictive significance of all independent variables (elevation, exposure, slope, ecological groups of plants, stone cover, grass cover, canopy cover, undergrowth and forest age) for the likelihood of the presence of each species. For all models, we started with a full model and followed a backward stepwise selection procedure to eliminate the effects that were furthest from statistical significance. Only the final models were presented. The categorical variables Habitat and Dominant tree were excluded from the statistical analysis because they consisted of too many levels while there were too few cases relative to the number of levels.

Spearman rank order correlation test was used for searching correlations among nest density of each species and the predictor variables used in the logistic regression analysis (missing data were pairwise deleted). All statistical analyses, except circular-linear correlation tests between exposure and ant species nest density, were performed using JASP 0.14.1. The circular-linear correlation tests and graphs were conducted with ORIANA 3.21.

A total of 256 mounds (active and abandoned) were found in 10 of the 11 studied sites and 104 (61%) of the 172 sampling transects (see Suppl. material 2: Table S1). The 229 active nests (89%) belonged to four RWA species: F. lugubris, F. rufa, F. pratensis and F. polyctena x rufa (Table 1). Abandoned nests were 27 (11%).

The most abundant species was Formica lugubris with 100 nests (44%), followed by F. rufa with 91 nests (40%), F. pratensis with 35 nests (15%) and the hybrid F. polyctena x rufa with 3 nests (1%). The average nest density of the four species was 11.1 nests/ha. The distribution of the nests is presented on the potential habitats’ maps for F. pratensis (Fig. 2A), F. rufa (Fig. 2B) and F. lugubris (Fig. 2C).

Figure 2. 

Potential habitats of RWA in Bulgaria. Localities (transects) recorded in the course of the present study are designated (red dots), previous literature data (squares) A Formica pratensis B Formica rufa C Formica lugubris.

The number of the recorded ant nests varied across the studied mountain ranges (Table 1). In Pirin Mt the greatest number of nests (63) was found, followed by Western Rhodopes (47) and Rila (35). In the remaining regions, between 0 and 18 nests were registered in each.

The most abundant population of F. lugubris was found in Pirin (43 nests, 14 nests/ha) and Western Rhodope Mts (35 nests, 12 nests/ha); F. rufa was most abundant in Belasitsa (13 nests and 13 nests/ha), Pirin (17 nests, about 5.5 nests/ha), Western Rhodope Mts (11 nests and 3.7 nests/ha), Rila and Osogovo (15 nests and about 3 nests/ha per each); F. pratensis – in Eastern Stara Planina (10 nests, about 5 nests/ha) and Rila (9 nests, about 2 nests/ha). Four of the 91 fenced nests of F. lugubris in Smolyan Forestry (Western Rhodope Mts) were found (Fig. 3).

Figure 3. 

One of the four active Formica lugubris nests fenced 50 years ago.

Formica aquilonia and F. truncorum were not found in our samplings during the survey. Similar to RWA Formica (Coptoformica) exsecta and Formica (Raptiformica) sanguinea were recorded in samples from the same monitoring sites.

The studied species had specific altitudinal distribution, though partly overlapping. Formica pratensis was found between 320 and 1173 m, F. rufa between 537 and 1650 m, and F. lugubris between 1040 and 2240 m. The elevation was a statistically significant predictor of the presence of the three RWA, with a higher probability for F. pratensis to be found at a lower elevation, and for F. rufa and F. lugubris to be present at higher elevations (Table 2–4; Fig. 4A).

Figure 4. 

Distribution of RWA by studied variables in percentage by species (F. pratensis: N = 35, F. rufa: N = 91, F. lugubris: N = 100) A elevation B slope C ecological groups of plants D stone cover E grass cover F canopy cover G dominant tree H forest age (transects with unknown values are excluded). For underlined species the impact is statistically significant.

Exposure was significantly correlated with the nest density of F. lugubris (Table 5) and showed a bidirectional (axial) distribution. The highest nest density corresponded to either NE-E or SW-W nest exposure (Fig. 5). Since the exposure of F. lugubris nests was also significantly correlated with elevation, with NE-E exposure corresponding to higher elevations and SW-W exposure to lower elevations (r = 0.401, p = 0.006, n = 34), we suppose that our results reflect rather an interrelationship of exposure with elevation and, perhaps, some other environmental parameters than exposure by itself.

Figure 5. 

A histogram of F. lugubris nest density and exposure. The frequency distribution of the nest number recorded at any particular direction is represented by different colours. For better visualisation, zero values are excluded.

The slope was a marginally significant predictor of F. pratensis and F. lugubris presence (Table 2, 4; Fig. 4B).

Along a moisture gradient, the ecological groups of plants were significantly correlated with the nest density of F. lugubris (Table 5). Most of its nests were found where mesophytes (53%) and meso-hygrophytes (35%) were dominant plants, and nests were rarely recorded in places where meso-xerophytes and xerophytes dominated (Fig. 4C).

All RWA species were most common at the lowest (0–5%) level of stone cover, however, this variable was a significant negative predictor of F. rufa presence only, and was negatively correlated with its nest density (Tables 3, 5; Fig. 4D).

Most of the nests of examined RWA species were found in habitats with 75–100% grass cover. This variable significantly predicted the presence of F. pratensis only and correlated positively with its nest density (Tables 2, 5; Fig. 4E).

Canopy cover negatively predicted the presence of F. lugubris (Table 4). Additionally, canopy cover was negatively correlated with the nest density of F. pratensis and positively correlated with the nest density of F. rufa (Table 5; Fig. 4F). Formica pratensis preferred habitats with canopy cover from 0 up to 60% and was lacking in habitats with canopy cover over 80%. Formica rufa had peaks of nest numbers between 40 and 80% of the canopy cover as the nests were situated on light spots in forests’ interiors. The majority of nests of F. lugubris (42%) were found at 40–60% and 21% of them at 0–20% canopy cover.

A great percentage (37%) of the nests of all studied species were found in the habitats of Picea abies (Fig. 4G). No RWA were found in Castanea sativa forests. In habitats with Juniperus sp., Pinus peuce, P. heldreichii and P. mugo we met only F. lugubris. In total, 65% of the nests of F. rufa were found within or in the vicinity of P. sylvestris forests, mixed forests with Fagus spp. or deciduous forests of Fagus spp.

Concerning the forest age, there was a significant positive correlation with the nest density of F. rufa (Table 5). This species was met even in forests older than 160 years, but the highest density was recorded in forests between 70 and 100 years old (Fig. 4H).

None of the species showed a statistically significant correlation with undergrowth density.

Most of the known localities of the particular RWA species (F. lugubris, F. pratensis and F. rufa) were confirmed in our study. The new findings are marked in the Suppl. material 2: Table S1. In all monitoring sites (except Eastern Rhodope Mts), at least one RWA species has been found. Formica lugubris was expected (Lapeva-Gjonova et al. 2010) but not found in the transects of Belasitsa and Osogovska Mts. It should be searched at a higher elevation. Formica rufa was also expected in Eastern Rhodope Mts and Strandza Mt. In Eastern Rhodope Mts the species should be searched for in the southern parts of the mountain. For Strandza Mt, it has been reported by Bobev (1972); Vatov and Bobev (1976); Wesselinoff (1979) and Vesselinov (1981) and should be searched also at a higher elevation. Formica pratensis was not found in our sampling transects on Pirin, Vitosha and Belasitsa Mt but mounds were found in all these mountains out of the transects.

Formica polyctena was not recorded in the course of the present study. Bobev (1972), Keremidchiev et al. (1972) and Vatov and Bobev (1976) reported also that F. polyctena is absent from the samples collected during the National Inventory of RWA (1970–1973). According to Bernhard Seifert (pers. comm. 2015), old data reported for F. polyctena from Bulgaria (Seifert 2008) are “a misplacement”. The three nests similar to F. polyctena, found in this research, were identified as F. polyctena x rufa and later confirmed by B. Seifert. Stable F. polyctena x rufa hybrid populations in Europe are known (Seifert 2018, 2021). There is no other recently published record on F. polyctena from Bulgaria either. So, we could accept Seifert’s opinion about the old findings: “The earlier findings of F. polyctena in Bulgaria, some 50 years ago, I suppose more likely to represent a misidentification rather than indicating a dying out process”.

Formica aquilonia was not found during our survey. If this species occurs in the country as reported for Rila Mt. at Zavrachitsa hut (Wesselinoff 1973) and Western Predbalkan at Belogradchik (Atanassov and Dlusskij 1992), its presence is probably very scarce. The samples identified as “F. aquilonia”, collected in 1982 from Bulgaria and preserved in the Senckenberg Museum of Natural History Görlitz, were kindly checked by B. Seifert in 2015 with modern taxonomic methods and were found to be seta-reduced F. lugubris (Pirin Mt) and seta-reduced F. pratensis (Rhodope Mts).

The average nest density (11.1 nests/ha) is greatly increased compared to that calculated at the National Inventory 1970–1973 (0.1 nests/ ha) given by Bobev (1972, 1973) and Vatov and Bobev (1976) (see Table 6). The reasons are the different sampling methodology, the difference in the calculation of the nest density, and the concentration of efforts for the studied species in the most suitable habitats during the fieldwork of this project.

Regarding the field’s methodology of the Inventory, there is scarce information in the Bulgarian State Archives and it is difficult to compare the results with confidence. Nevertheless, the confirmation of the localities and the new findings are extremely valuable from the conservation viewpoint.

As seen from the comparative Table 6, Western Rhodope Mts were the richest of the RWA region in 1970 and still, they are. The majority of the nests in the 1970s and nowadays are of F. lugubris and F. rufa. In our study, F. lugubris was the dominant species in Pirin Mt. Middle rich/poor regions in both projects remain Western and Central Stara Planina, Rila and Belasitsa Mts. There the species’ proportions remained approximately the same, although F. lugubris decreased at the expense of F. rufa and F. pratensis. In the poor regions (Strandza and Eastern Rhodope Mts) the dominance of F. pratensis was confirmed again in our study.

The nest density of F. rufa and F. lugubris in our survey is many times larger in the forest boundary area (in a strip with a width of up to 10 m) or small-size forest fragments with respectively higher solar radiation. Similar are the observations made by Punttila et al. (1994), Punttila (1996), Underwood and Fisher (2006), Babik et al. (2009), Crist (2009) and Chen and Robinson (2014) for RWA in the cool temperate forests in Europe. The density of the Formica rufa group is higher in deforested strips as they prefer nesting in sun-warmed places and use the forest interior as foraging area (Babik et al. 2009). In Western Poland, the nest density decreases towards the centre of clear-cuts although the influencing factors are not clear (Żmihorski 2010). The mounds in young forests and clear-cuttings are smaller and flatter, caused by the splitting of the large mounds into smaller colonies (Rosengren and Pamilo 1978; Domisch et al. 2005). During our survey, nests of Formica fusca (the host species of RWA) were often found in grassy runs or shrubs at the forest edges.

In our study, F. lugubris, F. rufa and F. pratensis were found at altitudes corresponding to the previously reported altitude from Bulgaria by Atanassov and Dlusskij (1992). In Switzerland, the great abundance of F. lugubris was at about 1000 m (Cherix et al. 2012) while in Bulgaria was higher, between 1500 and 1800 m. In Switzerland, F. rufa and F. polyctena were found at lower elevations up to 800 m (Vandegehuchte et al. 2017) similar to their optimal range in Bulgaria. Formica polyctena x rufa hybrid was found between 1200 and 1410 m, as the elevation preference for F. polyctena was up to 1200 m (Atanassov and Dlusskij 1992). In other European countries, almost every species of RWA occurs at lower altitudes than in Bulgarian mountains, which is the expected phenomenon of shifting up of altitudinal preferences of mountainous species in the southern part of their ranges.

In the present study, the nest density of F. pratensis and F. rufa was significantly influenced by the canopy cover. There was an example of shading effect: Wesselinov (1968) published an article about the high diversity of RWA for Parangalitza Biosphere Reserve in the Rila National Park. As a reserve since 1933 (Darzhaven vestnik 1933) with about 300–400 years old trees of Picea abies, the wood clearance, removing the old fallen woods and felling had been strongly forbidden. The consequences nowadays are that as the forest interior is almost closed (over 80% canopy cover), the RWA are only to be found in the vicinity of the forest, around the roads and the alpine meadows at present. A similar situation was observed in other old reserves and RWA populations there were decreasing. Our observations confirmed the conclusions of other authors (Collingwood 1979; Punttila et al. 1994; Dekoninck et al. 2010; Tsikas et al. 2016; Serttaş 2020) that RWA populations decline with increasing the canopy closure and a little intervention as forest thinning for light spots would be a support for the RWA existence.

In Bulgaria, the mounds of F. lugubris were mostly located on north-east-facing slopes, similarly to the same species in Switzerland, found mainly at an eastern aspect (Freitag et al. 2016b; Vandegehuchte et al. 2017). In southern Norway, it was strongly orientated to the South (Hill et al. 2018).

The positive correlation of grass cover to both F. pratensis presence and nest density was expected because this species is an open habitat specialist (Seifert 2018). This variable did not show a statistically significant impact on the abundance of other RWA as well as in West Poland for F. rufa and F. polyctena (Żmihorski 2010). According to Sorvari (2016), the higher ground vegetation benefits RWA as it plays a protection and feeding role.

The dominant tree species associations, where the RWA species occur, were Picea abies, Pinus sylvestris, Abies alba, young Juniperus spp. communities (according to Atanasov and Dlusskji 1992) and also Pinus mugo, P. heldreichii and P. peuce communities (according to our study).

The conservation needs of the RWA are underestimated (Sorvari 2016). For a viable colony, these territorial species need at least a few hectares of stable habitat with enough food resources to maintain sexual forms for reproduction. The loss, shading, drying/flooding, disturbance or destruction of their habitat are their major threats (Dekoninck 2010). Not only human activities but changes in climatic conditions and natural enemies (as pathogens) may cause a decline of a colony. As RWA play a keystone role in the ecosystems (Dlussky 1967), the reduction or extinction of their populations may have a huge effect on the local and even global ecosystems. Therefore, studying the most suitable habitat characteristics in detail, and conducting regular monitoring studies on RWA populations are of primary importance.